Generating Behaviorally Diverse Policies with Latent Diffusion Models
This addresses a scalability problem for researchers and practitioners in reinforcement learning by enabling efficient storage and flexible behavior selection, though it is incremental as it builds on existing QD-RL and diffusion model techniques.
The paper tackles the high space-complexity and poor scalability of storing thousands of policies in Quality Diversity Reinforcement Learning by proposing a method to distill them into a single generative model using diffusion models, achieving a 13x compression ratio while recovering 98% of original rewards and 89% of original coverage.
Recent progress in Quality Diversity Reinforcement Learning (QD-RL) has enabled learning a collection of behaviorally diverse, high performing policies. However, these methods typically involve storing thousands of policies, which results in high space-complexity and poor scaling to additional behaviors. Condensing the archive into a single model while retaining the performance and coverage of the original collection of policies has proved challenging. In this work, we propose using diffusion models to distill the archive into a single generative model over policy parameters. We show that our method achieves a compression ratio of 13x while recovering 98% of the original rewards and 89% of the original coverage. Further, the conditioning mechanism of diffusion models allows for flexibly selecting and sequencing behaviors, including using language. Project website: https://sites.google.com/view/policydiffusion/home